A Study of Biological Sequence Classification Based on Local Globally Weighted KNN Algorithm

Classification and identification is of great significance in biological sequence analysis research, which can help to solve the problems of functional identification, medical diagnosis, epidemic monitoring and so on. At present, there are mainly machine learning methods and non-machine learning methods to classify DNA sequences, and both methods have their own merits when the data volume is small, but with the explosive growth of genomic data, the advantages of machine learning methods become more and more obvious. Meanwhile, compared with other machine learning methods, there are problems such as complex training models, overfitting, and error amplification when dealing with sequence classification problems. To address these problems, the LGWKNN algorithm based on the interaction of local weights and global weights is introduced, which experimentally confirms the feasibility of sequence classification as well as the advantages over traditional KNN algorithms and weighting algorithms, achieving a classification accuracy of 97.71%.